Eurekah Bioscience Collection

Transcription

1 in Malignant 5/31/06 12:09 PM in Malignant to Leukemia and Lymphoma Eurekah Bioscience Collection in Malignant Sarah E. enrickson Elena M. artmann German Ott Andreas Rosenwald* The practice of clinical medicine and the process of biomedical research have been transformed by the decoding of the human genome. The use of DNA microarrays to find gene expression patterns in disease and biological processes has already begun to have a significant impact on modern medicine. The study of hematological malignancies has particularly benefited from gene expression profiling, including discoveries about prognosis, mechanism and efficacious choice of therapeutic regimens. DNA microarrays have led to the discovery of better prognostic tools, including the use of Zap-70 in B-Cell Chronic Lymphocytic Leukemia (B-CLL) as an indicator of worse prognosis. Studies of Diffuse Large B cell Lymphoma (DLBCL) have defined two molecular subgroups, with significantly different mortality rates and responses to conventional therapy. In Follicular Lymphoma (FL), the variable clinical course could be associated with molecular signatures reflecting a possible interaction between tumor cells and infiltrating immune cells. The molecular mechanisms of Mantle Cell Lymphoma (MCL) have also begun to be clarified, with a more detailed understanding of the roles of cell cycle and DNA damage pathways that are responsible for the varying degree of tumor cell proliferation and different clinical outcome in this lymphoma. While important discoveries have been made in leukemias, lymphomas and many other cancer subtypes using gene expression profiling, there are many questions left to study and the translation of these tools and their results into the clinic has just begun. Page 1 of 1

2 5/31/06 12:09 PM in Malignant to Leukemia and Lymphoma Eurekah Bioscience Collection in Malignant With the development of gene expression analysis at the genomic scale comes the possibility of accurately stratifying cancer patients by the molecular characteristics of their tumors and the development of individualized, tumor-specific therapy. We currently use histological examination of cancer specimens, clinical characteristics of patients and, more recently, genetic information to classify tumors into pathological entities and to stratify cancer patients into treatment paradigms. For example, the current diagnosis of follicular lymphoma (FL) brings together morphological aspects of the tumor cell infiltrate (atypical follicular structures), immunophenotyping of the tumor cells (e.g., coexpression of the markers CD10 and BCL2) and cytogenetics (translocation of the BCL2 oncogene). 1 Gene expression profiling provides the potential to gain a deeper understanding of the complex biological and molecular basis of lymphomas, facilitating discovery of new drug targets and new therapeutic regimens. In addition, while the most basic goals of diagnosis in the clinical setting are to clarify for the patient and the care provider both a clear prognosis and a therapeutic plan, our current standard of diagnosis does not reliably yield either. Within diagnostic categories there can be great variation among patients both in their response to therapeutic regimens and their overall survival. This situation provides the opportunity for molecular diagnostics to be developed which may help to provide clear definitions of various forms of cancer, and allow more accurate diagnosis and more applicable therapy. This will hopefully lead to more effective outcome prediction for lymphoma patients, for whom there is currently considerable variation in their clinical course even within one single lymphoma entity. Currently, accurate prognostic markers for use at the time of diagnosis as well as therapeutic alternatives are lacking in many lymphomas. While high-throughput genomic techniques are applicable to all facets of human pathology, lymphomas will be the specific focus of this chapter. The use in clinical practice is not the only goal of gene expression profiling. Careful characterization of the patterns of gene expression of cancer cells versus normal cells can clarify the cells of origin of these conditions and the pathways that are altered during neoplastic transformation. With a clearer understanding of the mechanisms of transformation of normal cells into cancerous cells, it will be possible to design more targeted therapies with greater efficacy and fewer side effects. *Corresponding Author: Andreas Rosenwald-Institute of Pathology, University of Würzburg, Germany, Rosenwald@mail.uni-wuerzburg.de Page 1 of 1

3 5/31/06 12:09 PM in Malignant to Leukemia and Lymphoma Eurekah Bioscience Collection in Malignant Gene expression profiling is based on the use of DNA microarrays to assess the level of gene expression at a given time in a population of cells. There are two main forms of microarrays. 2 The first involves the use of robotic equipment to spot complementary DNAs (cdnas) onto glass slides coated to bind DNA effectively. 3, 4 The cdnas are attached to the slides in carefully planned grids, which allows accurate analysis of sample binding patterns after experiments take place. In order to use these arrays, the cell sample of choice is lysed, mrna is extracted and reverse transcribed to cdna which is fluorescently labeled. These fluorescently labeled probes are allowed to anneal to the cdna microarray and the degree of hybridization is assessed by fluorescent microscopy. An example of this type of DNA microarray is the Lymphochip which contains cdnas that are thought to be involved with the initiation and pathogenesis of leukemia and lymphoma, as well as genes thought to be generally involved in immune function. 3 This platform was created before DNA microarrays were widely available commercially. The genes on the Lymphochip were chosen by sequencing cdna from libraries of germinal center B cells, leukemias and lymphomas, yielding 15,000 expressed sequence tags (ESTs) enriched in immune cells. 5 Approximately 3500 additional genes were added to this pool based on their known roles in immune system function. A second type of DNA microarray is that provided by Affymetrix. This technique uses multiple representative oligonucleotides for each gene of interest which are synthesized directly onto silicon wafers. In addition, for each oligonucleotide there is a mismatch probe arrayed on the chip as well, with a slightly altered sequence. Samples are prepared in a similar manner as for the cdna arrays, although there is one round of RNA amplification added in the Affymetrix protocol. Pairing of sample cdna to the perfect match oligonucleotide alone (with no binding to the mismatch sequence) is considered evidence of gene expression in that sample. Page 1 of 1

4 in Malignant to Leukemia and Lymphoma Eurekah Bioscience Collection in Malignant The quantity of data generated using these techniques required the development of new analytical techniques to find the patterns and alterations of relevant pathways among all the less important changes in gene expression (noise). There are two main paradigms of analysis that have emerged. The first is based on the idea that the answer should come from the data itself, without any input from the researcher, which is known as an unsupervised approach. The preferred method of analysis in this case is often hierarchical clustering. 6 This technique uses an algorithm to cluster together genes with correlated expression or samples with similar gene expression patterns (or both) simply based upon the intrinsic relationships among gene expression patterns. An example of a different approach is the search for genes that could differentiate between predefined cancer or lymphoma subgroups, e.g., between immunoglobulin heavy chain (IgV) -mutated and -unmutated B cell chronic lymphocytic leukemia (B-CLL) patients (see below). 7 This method of analysis is based on the concept that the data should follow known biological principles and is generally regarded as supervised analysis. 8 In this case, a model is designed to distinguish between a chosen number of groups, rather than being allowed to discover as many groups as it finds in the data. 9 An example of unsupervised analysis is the discovery of two subgroups of diffuse large B-cell lymphomas (DLBCL), termed activated B cell-like (ABC) and germinal center B cell-like (GCB) DLBCL that are biologically distinct and associated with a different clinical course. 10 Once genes are clustered, based on similarity in gene expression, we can identify gene expression signatures that represent sets of genes which all participate in a biological process or characterize a given cell type. 11, 12 For example, signatures can be derived for nontransformed B and T cells, allowing the identification of novel or unrecognized genes involved in these immune cells by their coordinate expression with genes known to be expressed by these cells (i.e., TCR, TCR and CD3 in T cells). 12 In addition, this allows us to derive biological meaning from the genomic data by identifying relevant cell types and important pathways in the lymphoma samples. For example, in the studies of DLBCL patients, the proliferation signature was found to be associated with adverse clinical outcome, whereas expression of the lymph node signature which is composed of genes mostly derived from nonmalignant bystander cells in the lymphoma specimens confers a more 11 Page 1 of 2

5 favorable clinical course. Recognition of these gene signatures allows a better understanding of the biological processes and pathways involved in a given sample, and fosters development of prognosis prediction algorithms. Representative genes from each of the prognostically informative signatures can be combined into prognostic predictors that can be used to classify patients into prognostic and therapeutic groups. Page 2 of 2

6 to Leukemia and Lymphoma in Malignant to Leukemia and Lymphoma Figures Figure 1. Figure 2. Figure 3. Figure 4. Eurekah Bioscience Collection in Malignant to Leukemia and Lymphoma Diffuse Large B-Cell Lymphoma (DLBCL) Diffuse large B cell lymphoma is the most common lymphoma in adults. While many therapeutic regimens have been attempted, the previous standard of care, COP (an anthracycline based chemotherapeutic regimen) was able to cure DLBCL in only 35-40% of patients. 13 Attempts to improve survival with multiple alternate regimes did not make significant changes in that success rate. 14 owever, the addition of Rituximab (a monoclonal anti-cd20 antibody) to the traditional COP regimen did lead to a significant increase in survival, 15 although prognosis for any individual patient remains challenging to predict. It was therefore hypothesized that the failure to improve DLBCL cure rates may reflect the existence of multiple subgroups of patients, each with a slightly different pathogenic mechanism. This would imply that one particular therapeutic regimen may not be equally effective among different DLBCL subgroups. A number of methods exist for stratifying DLBCL patients in order to predict outcome, including the currently used International Prognostic Index (IPI) which focuses on clinical parameters that include: Age of the patient, Eastern Cooperative Group (ECOG) performance status, tumor stage, lactate dehydrogenase level, and the number of sites of extranodal disease. 16 owever, the IPI is not able to stratify patients into different therapeutic regimens effectively. 13 In order to determine whether or not DLBCL is a single condition, which simply required discovery of a novel therapeutic regimen, or multiple conditions each of which required different therapies, gene expression profiling was undertaken. The initial study by Alizadeh and colleagues found two groups of DLBCL patients by hierarchical clustering. 10 In this study, gene expression patterns from the patient samples were compared to gene expression patterns from normal B lymphocyte subsets, allowing parallels to be drawn between the lymphoma samples and possible cells of origin. The first patient subgroup, which was characterized by germinal center B-cell-like (GCB) gene expression patterns, had a better prognosis with COP-based therapy than the second group, which was characterized by activated B-cell-like (ABC) gene expression patterns. A followup study from the same group used a larger patient cohort and confirmed the existence of GCB- and ABC-like subsets of DLBCL. In addition, a third DLBCL subgroup (Unclassified DLBCL) was detected that did not show characteristic gene expression patterns of GCB or ABC DLBCL. 11 Importantly, these subgroups were independent of the IPI and only partially correlated with the Page 1 of 8

7 to Leukemia and Lymphoma histological diagnosis with centroblastic monomorphic cases being more common among GCB DLBCL, while centroblastic polymorphic and immunoblastic cases were more frequently observed in ABC DLBCL. The two major DLBCL groups each have defining characteristics beyond clustering on DNA microarrays ( Fig. 1). The germinal center B cell-like designation is prognostically favorable (five year survival of 60% versus 35% for ABC DLBCL). 11 This group was also found to more commonly have Bcl-2 translocations and c-rel amplifications than ABC DLBCL. 11 Additional differences in genetic alterations include more frequent chromosomal gains of 12q in GCB DLBCL, whereas genomic gains in 3q and 18q are predominantly found in ABC DLBCL (unpublished data). In contrast, only in cell lines representing the ABC DLBCL subgroup there was constitutive NF- B expression. 17 NF- B is regulated by the I B family, which retains NF- B in the cytoplasm until I B kinase (IKK) phosphorylates I Bs. This targets IkBs for ubiquitin mediated proteasomal degradation, thus freeing NF- B and allowing it to translocate to the nucleus and initiate target gene transcription. NF-kB is important for normal B cell development and survival, and constitutively nuclear NF- B has been implicated in a number of cancers. The association of constitutive NF- B expression in the ABC subgroup of DLBCL patients is an encouraging finding, in that it provides a novel target for drug development in the DLBCL subtype with the poorest prognosis. Interestingly, PS-341 (Velcade, Bortezomib), a proteasomal inhibitor that targets NF-kB by preventing degradation of its inhibitor, I B, has been shown to have efficacy in treating multiple myeloma and is in trials for DLBCL patients. The microarray data allowed the diagnosis of DLBCL to go far beyond simply stratifying patients into three subgroups. A molecular predictor of survival after chemotherapy for DLBCL was created based on outcome data correlation with gene expression patterns. 11 The goal of this analysis was the identification of genes correlated with outcome, based on a Cox proportional hazards model. Genes were classified into so-called gene expression signatures that had been previously defined; genes belonging to the germinal center B cell signature, lymph node signature and MC class II signature were associated with good outcome in DLBCL patients, while the proliferation signature was associated with poor outcome. As expected, the expression of the germinal center B cell signature was high in the GCB DLBCL subgroup, whereas the proliferation signature was generally more highly expressed in the ABC DLBCL subgroup. At the average, the MC class II signature was similarly expressed in all three DLBCL groups. For the development of a gene expression-based outcome predictor, 17 genes were selected which were highly variable in expression and which represented the expression level of the respective signatures. Overall, the molecular predictor was found to be independent of the IPI. Using Affymetrix arrays, Shipp and colleagues also studied DLBCL patients and developed a gene expression outcome predictor that included, among other genes, the expression of NOR1, PDE4B and PKC-, which are all involved in 20 apoptotic pathways. The distinction between GCB and ABC DLBCL was also evident in this data set and the DLBCL subgroups had different survival. Interestingly, Shipp et al had showed in their dataset that PDE4 was Page 2 of 8

8 to Leukemia and Lymphoma overexpressed in samples from DLBCL patients with poor prognosis. They put forward a hypothesis that the resulting inhibition of camp would prevent apoptosis in these cells. Subsequently, PDE4 was confirmed as a target in poor prognosis DLBCLs and the apoptotic pathway blocked in DLBCL was found to be dependent on the PI3K/AKT pathway. 21 A very recent study from the same group has yielded three DLBCL subgroups by microarray profiling, with one characterized by a higher level host response, including immune cell infiltration of the tumors. 22 The importance of the tumor microenvironment in DLBCL has also recently been shown in follicular lymphoma (FL), 23 which may foreshadow more emphasis on the interaction between the lymphoma cells and the host environment in lymphoid neoplasms. The studies by both the Shipp and Staudt groups have led to the identification of potential therapeutic targets in different subgroups of DLBCL. The Staudt group proposed the NF B pathway as a target for ABC DLBCL. 17 The Shipp group is interested in finding inhibitors of PDE4 and the PI3K/AKT pathway as well as PKC, and clinical trials with a PKC- inhibitor are ongoing (M. Shipp, personal communication). B-Cell Chronic Lymphocytic Leukemia (B-CLL) B-cell chronic lymphocytic leukemia (B-CLL) is the most common leukemia in the Western hemisphere. 24 owever, while B-CLL is often diagnosed at an early stage of the disease, patients can either develop indolent disease or may suffer an acute, precipitous decline. Are there truly two separate types of B-CLL with separate prognostic markers and outcomes? In order to study this question, it was necessary to find prognostic markers that identified which outcome was most likely for each patient. Initially, cytogenetic differences were studied to find prognostic markers in B-CLL, and 17p or 11q deletions were suggested to be predictors of poor outcome. 25, 26 In two independent studies, however, the presence of somatic mutations in the immunoglobulin heavy chain variable regions (IgV ) of the tumor cells was found to be a predictor of better patient outcome. While this finding was a landmark discovery, it is not suitable as a practical clinical test since it is expensive and time consuming to generate IgV sequences on all B-CLL patients, and many clinical laboratories may not have the capacity to do this test on a routine basis. In addition, it is unclear what threshold to use for differentiating IgV -mutated from IgV - unmutated B-CLL cases (thresholds from 96-98% have been used). CD38 was suggested as a surrogate marker for the IgV mutation status; however, while CD38 expression is of prognostic significance, it failed to be confirmed as a useful surrogate marker for the IgV mutation status in two large studies , 28 24, 27 The finding of IgV somatic mutation levels having prognostic value might have implied that there were truly two different subtypes of B-CLL, each with their Page 3 of 8

9 to Leukemia and Lymphoma own progenitor cell, with a pregerminal B cell in the case of IgV -unmutated B- CLL and a post-germinal B cell in the case of IgV -mutated B-CLL. owever, 7, 29 two separate DNA microarray studies showed clearly that IgV- mutated and -unmutated B-CLLs share a homogenous gene expression signature that allows B-CLL as a whole to be distinguished from other leukemias and lymphomas. Thus, B-CLL, regardless of IgV somatic mutation frequency, has a distinct transcriptional profile and therefore seems to constitute one single disease. The IgV mutation status therefore remained an important prognostic factor, and many groups began using the DNA microarray data to find a gene expression pattern that could serve as a proxy. While B-CLL has a generally homogenous gene expression signature across IgV mutated and unmutated samples, the data from gene expression studies also showed a small number of genes that had different expression levels between patients with mutated and unmutated IgV genes 7 ( Fig. 2). The best correlate for an unmutated IgV status was found to be ZAP-70, a tyrosine kinase previously only known for its role in T cell receptor signaling. Interestingly, ZAP-70 was shown to be expressed at negligible levels in IgV mutated B-CLL cells, while it was expressed at significant levels in IgV 7, 30, 31 unmutated B-CLL cells. A large study in the UK further validated the correlation between ZAP-70 expression and the IgV mutation status. 32 Other groups have validated ZAP70 as a prognostic indicator as well, and recently ZAP-70 has been shown to be a more useful predictor of need for therapy in B- CLL than the IgV mutation status. 34 In order to facilitate transfer from bench to bedside, both RT-PCR and immunohistochemical methods were developed to measure ZAP-70 levels. The use of immunohistochemistry allows proxy assessment of the IgV status without separating out the tumor cells, by using a concomitant stain for the B cell marker CD19. The most promising clinical application, however, may be the measurement of ZAP-70 expression by flow cytometry analysis, since this technique is widely used in the standard work-up procedure of B-CLL samples in many laboratories. 33, 34 owever, many issues remain to be completely resolved, including the exact level of ZAP-70 expression necessary to be considered positive in the various testing modalities, as well as the percentage of cells that must each be ZAP-70 positive (since there can be heterogeneity among tumor cells) and research on these issues is currently underway. It has become clear recently that DNA damage response pathways play a role in response of B-CLL cells to therapy. There is an intriguing subset of B-CLL cases that have a shared alteration in their response to DNA damage, specifically due to ATM and p53 mutations As expected, with defects in DNA damage repair, these B-CLL cells have a different response to ionizing radiation than B- CLL samples with unmutated DNA damage repair genes. 38 Specifically, B-CLL cells with mutations in p53 and ATM failed to upregulate pro-apoptotic target genes (like FAS and TRAIL-receptor 2) suggesting that restoration of these mutated genes in the subset of patients with p53 and ATM mutations may have therapeutic effects Page 4 of 8

10 to Leukemia and Lymphoma This is especially relevant when thinking about the current therapeutic options for symptomatic B-CLL. Purine analogs, like fludarabine, act as anti-neoplastic agents by inducing double strand breaks in DNA, leading to programmed cell death. In a recent microarray study of gene expression in previously untreated B- CLL patients at 3 hours, 6 hours, and days 2, 3, 4 and 5 after initiation of fludarabine therapy, it became clear that the major alteration in gene expression was the initiation of the p53 pathway. Since there were no other clear changes in gene expression, this pathway may be the major (or exclusive) mediator of drug function. 39 This may lead to in vivo selection of p53 mutant subclones, leading to the development of drug resistant disease due to therapy. It will be important to conduct further studies on this phenomenon, including an analysis of alternate therapies and their likelihood of selecting for similar mutant B-CLL subclones. Mantle Cell Lymphoma (MCL) Mantle cell lymphoma (MCL) is a mature B-cell lymphoma that makes up 6% of all B-cell non-odgkin's lymphomas (B-NL). 1 MCL has a median patient survival of three to four years, and while survival is heterogeneous, there is generally an aggressive clinical course with poor response to chemotherapy. The classic translocation associated with this condition is the t(11;14)(q13;q32) which leads to cyclin D1 overexpression and effects at the G1/S checkpoint of the cell cycle. 40 MCL is also associated with ATM inactivation and p53 mutations. Previous attempts to stratify MCL patients have identified characteristics associated with poor survival including a blastic morphological variant, increased tumor cell proliferation, INK4a/ARF locus deletion and p53 mutation or protein overexpression, but none of these biological features have been used to successfully stratify MCL patients into therapeutic categories. 40 In a large gene expression profiling study, DNA microarrays of MCL patient samples were used to derive molecular prognostic information. 41 In this study, gene expression patterns of Cyclin D1-positive MCL specimens were compared to other B-NL subsets and a large set of genes was derived that is characteristically expressed at high levels in MCL. Moreover, a small subset of MCL-like lymphoma cases were studied that show morphologic and immunophenotypic characteristics, but lack expression of Cyclin D1. A substantial proportion of these cases showed an expression profile identical to Cyclin D1-positive MCL cases and, therefore, these cases may represent a small subgroup of bona fide MCL that lack expression of Cyclin D1. As a second step, a predictor of patient survival was constructed using the same gene expression array data. This predictor revealed that the proliferation signature was correlated with poor outcome in MCL. The predictor was optimized with a randomly selected training set of patient samples and then validated with a randomly selected validation set of patient samples. Overall, the accurate quantitative measurement of proliferation in MCL cells, provided by proliferation signature genes, identified subsets of MCL patients that differ in their survival times by almost 6 years ( Fig. 3). Several molecular features were associated with increased proliferation in MCL cells. First, higher levels of Cyclin D1 expression were found to be associated with an increase of tumor cell proliferation. MCLs can express different Cyclin Page 5 of 8

11 to Leukemia and Lymphoma D1 mrna isoforms that differ in the lengths of their 3' untranslated regions (UTR). The 4.5kb version may have reduced stability due to a longer UTR and the presence of an RNA destabilizing element in this region. A shorter isoform of 1.7kb may be more stable due to lack of this destabilizing element. MCL cases with abundant expression of the short Cyclin D1 isoform had higher levels of overall Cyclin D1 mrna and, therefore, an increased rate of proliferation. Second, INK4a/ARF locus deletions were found in 18/85 of the cases by quantitative PCR, and more deletions were observed among the highly proliferative cases. Interestingly, BMI-1, a transcriptional repressor of the INK4a/ ARF locus was also highly expressed in a subset of highly proliferative MCL cases. In accordance with previously published data, both p53 and ATM deletions were also found in this set of MCL patients, but neither had a strong correlation with survival or proliferation. Interestingly, mathematical models including the level of Cyclin D1 expression or the INK4a/ ARF deletion status alone or in combination did not perform as well in predicting survival than the proliferation signature-based outcome model. Thus, the gene expression-based model may capture additional oncogenic events in MCL cells that are presently not known and serve as a global integrator of molecular alterations in MCL. 41 Follicular Lymphoma (FL) Follicular lymphoma (FL) is the second most common form of B-cell non- odgkin (B-NL). 1 FL is often associated with the chromosomal translocation t(14;18) that causes BCL-2 dysregulation, leading to reduced apoptosis. Clinically, there is a variable progression of the disease, from aggressive lymphoma to an indolent condition with intermittent episodes. At the present time, there are no robust biological markers available that predict the clinical course of FL patients at the time of diagnosis. In order to search for predictive markers and to identify the molecular basis of the biological and clinical heterogeneity of FL, gene expression analysis and survival signature analysis was undertaken, using RNA extracted from almost 200 FL patient samples 23 ( Fig. 4). Samples were split into training and validation sets, and a statistical survival model was developed using only the training set. In particular, a Cox proportional hazards model was used to find genes associated with survival in the patient population. ierarchical clustering was then applied to group single genes that were associated with favorable and poor survival into gene expression signatures, 12 and two of these signatures were found to have statistical synergy : One associated with good prognosis (immune response 1) and one associated with poor prognosis (immune response 2). Interestingly, both signatures were derived from nonmalignant bystander cells in the lymph node specimens and not from the tumor cells themselves. The signature associated with a more favorable clinical course (immune response 1) contains genes associated with subsets of T cells (e.g., CD8B1, ITK, STAT4). owever, the presence of this signature is not simply due to the number of infiltrating T cells in the tumor sample, since a number of pan-t cell markers were not part of this signature. The signature associated with poor prognosis (immune response 2) contains genes associated with macrophages and dendritic cells (e.g., TLR5, LGMN and C3AR1). This model therefore predicts that the Page 6 of 8

12 to Leukemia and Lymphoma relative levels of subsets of infiltrating cells are of prognostic significance in FL, which is intriguing for understanding the pathogenesis of this condition. Future studies in FL will likely place emphasis on investigating the interaction between the neoplastic B cells and the nonmalignant bystander cells which appears to be of importance for the biological and clinical behavior of this lymphoma subtype. A very recent study used microarray data to create a gene expression profile that could help differentiate between aggressive and indolent forms of FL, thus facilitating treatment decisions. 42 The profiles for indolent and aggressive clinical behaviour were created using sets of paired samples from indolent and aggressive stages of FL from the same patients. The final profiles allow stratification of samples into indolent and aggressive categories with a 93% success rate. This profile is not intended to project the transition from indolent to aggressive stages of the disease; it is rather a diagnostic tool to aid in therapeutic decisions. Further analysis of the genes that are more highly expressed in the aggressive form of FL may help elucidate the transition from indolent to aggressive disease. The Future of in Diagnosis, Prognosis and Therapeutic Choices Over the past six years, gene expression profiling has been used to create molecular profiles of various cancer subtypes. Lymphoid malignancies, especially the more common subtypes of B-NL, have been well-studied and gene expression profiling data have yielded many new insights into these conditions. The first goal was to clarify whether these diseases, with historically diverse prognostic outcomes, are truly single conditions or diseases with multiple subgroups that might each benefit from individualized therapy. For example, the ABC and GCB-like subgroups in DLBCL represent two lymphoma subgroups with both different molecular characteristics and different responses to standard therapy. An extension of this goal is the identification of single markers for clinical/prognostic subgroups that will allow stratification of patients quickly and without the need for microarray expertise. This has been seen in B-CLL, where the identification of ZAP-70 as a proxy for the mutational status of the immunoglobulin heavy chain variable region makes the discoveries from basic science laboratories applicable to routine clinical diagnostics. Another important goal of gene expression profiling is the identification of differentially implicated oncogenic pathways in the newly discovered subsets of lymphoid malignancies that could be targeted for future drug development. In this regard, ABC DLBCL was found to have constitutive NF B expression and subsets of DLBCL patients are characterized by activation of PKC ; exciting clinical trials, in which these potent oncogenic pathways are targeted by specific inhibitors, are ongoing. It has to be noted, however, that all gene expression studies summarized in this chapter were performed in a retrospective manner, and patients had not been treated according to currently used therapeutic regimens. Thus, the benefit of molecular diagnostic studies and, in particular, of gene expression profiling, will have to be tested in future multi-center clinical trials. Once sufficient data has been collected to define molecular signatures that allow the stratification of patients into subgroups with prognostic and therapeutic implications, gene expression may make its way into the mainstream of clinical practice as an adjunct diagnostic tool. There is clearly much research left to be done both in Page 7 of 8

13 to Leukemia and Lymphoma lymphoid malignancies and beyond, but the goal is clear. Acknowledgements Andreas Rosenwald is supported by the Interdisciplinary Center for Clinical Research (IZKF) of the University of Würzburg, Germany. Sarah enrickson is supported by the NI Medical Scientist Training Program. Page 8 of 8

14 Book eurekah 5/31/06 12:11 PM in Malignant to Leukemia and Lymphoma Eurekah Bioscience Collection in Malignant to Leukemia and Lymphoma Figure 1. Gene expression profiling identified molecular subgroups of diffuse large B-cell lymphoma (DLBCL). GCB and ABC DLBCL differ in their presumed cell of origin, in underlying genetic alterations and in their clinical behavior (for details see text). Page 1 of 1

15 Book eurekah 5/31/06 12:11 PM in Malignant to Leukemia and Lymphoma Eurekah Bioscience Collection in Malignant to Leukemia and Lymphoma Figure 2. Gene expression profiling identified the clinical marker ZAP-70 in B- CLL. ZAP-70 appears to correlate well with the mutational status of the immunoglobulin (Ig) heavy chain gene in B-CLL cells and predicts survival of B- CLL patients. Potential applications in the routine diagnostic setting involve ZAP-70 detection by immunohistochemistry or by flow cytometry. Page 1 of 1

16 Book eurekah 5/31/06 12:11 PM in Malignant to Leukemia and Lymphoma Eurekah Bioscience Collection in Malignant to Leukemia and Lymphoma Figure 3. Gene expression profiling identified a proliferation-associated signature as a strong predictor of survival in mantle cell lymphoma (MCL). The prognosis of patients with a high score of the expression-based proliferation signature is poor, while the prognosis is more favorable in patients with a low expression of the proliferation signature. Potential clinical application involves the measurement of proliferation-associated markers by immunohistochemical methods (such as staining for Ki 67). Page 1 of 1

17 Book eurekah 5/31/06 12:12 PM in Malignant to Leukemia and Lymphoma Eurekah Bioscience Collection in Malignant to Leukemia and Lymphoma Figure 4. Immune response signatures predict outcome in follicular lymphoma (FL). A mathematical predictor using Immune Response 1 and Immune Response 2 signature genes strongly predicts outcome of FL patients at the time of diagnosis (for details see text). Page 1 of 1